3 Experimental Error

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 precision
(reproducibility) of the measurement. If the results of measuring the same quantity by dif-
ferent methods agree with one another, 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

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 propaga-
tion of uncertainty in lab calculations.

3-1 Significant Figures

Significant figures: minimum number of digits The number of significant figures is the minimum number of digits needed to write a given
required to express a value in scientific value in scientific notation without loss of accuracy. The number 142.7 has four significant
notation without loss of accuracy figures, because it can be written 1.427 102. If you write 1.427 0 102, you imply that

3-1 Significant Figures 39

Figure 3-1 Scale of a Bausch and Lomb Percent transmittance
Spectronic 20 spectrophotometer. Percent 0 10 20 30 40 50 60 70 80 90 100
transmittance is a linear scale and
absorbance is a logarithmic scale.

∞ 2 1.0 0.5 0.4 0.3 0.2 0.1 0.05 0

Absorbance

you know the value of the digit after 7, which is not the case for the number 142.7. The num-
ber 1.427 0 102 has five significant figures.
The number 6.302 106 has four significant figures, because all four digits are neces-
sary. 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 end of
106 0.010 6 0.106 0.106 0 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 Spec-
tronic 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 significant 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 transmittance is near 58.3. Because the
transmittance scale is smaller than the absorbance scale at this point, there is more uncer-
tainty in the last digit of transmittance. A reasonable estimate of uncertainty might be
58.3
0.2. There are three significant figures in the number 58.3.
Interpolation: Estimate all readings to the When reading the scale of any apparatus, try to estimate to the nearest tenth of a divi-
nearest tenth of the distance between scale sion. On a 50-mL buret, which is graduated to 0.1 mL, read the level to the nearest 0.01 mL.
divisions. For a ruler calibrated in millimeters, estimate distances to the nearest 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 people, not 4.000
0.002
people!

3-2 Significant Figures in Arithmetic

We now consider how many digits to retain in the answer after you have performed arith-
metic 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.

40 CHAPTER 3 Experimental Error

 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 are Inspect the legend of the periodic table inside
limited by the least-certain one. For example, the molecular mass of KrF2 is known only to the cover of this book. Be sure you can
the third decimal place, because we only know the atomic mass of Kr to three decimal interpret uncertainties in atomic masses. For F
and Kr, the atomic masses are
places:
F: 18.998 403 2
0.000 000 5
18.998 403 2 (F) Kr: 83.798
0.002
 18.998 403 2 (F)
 83.798 (Kr)
121.794 806 4
123
Not significant

The number 121.794 806 4 should be rounded to 121.795 as the final answer.
When rounding off, look at all the digits beyond the last place desired. In the preceding Rules for rounding off numbers
example, the digits 806 4 lie beyond the last significant decimal place. Because this number
is more than halfway to the next higher digit, we round the 4 up to 5 (that is, we round up to
121.795 instead of down to 121.794). If the insignificant figures were less than halfway, we
would round down. For example, 121.794 3 is rounded to 121.794.
In the special case where the number is exactly halfway, round to the nearest even digit.
Thus, 43.55 is rounded to 43.6, if we can only have three significant figures. If we are retain-
ing 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 ration-
ale for rounding to an even digit is to avoid systematically increasing or decreasing results
through successive round-off errors. Half the round-offs will be up and half down.
In the addition or subtraction of numbers expressed in scientific notation, all numbers Addition and subtraction: Express all numbers
should first be expressed with the same exponent: with the same exponent and align all numbers
with respect to the decimal point. Round off
1.632 105 the answer according to the number of
1.632 105
 0.041 07 105 decimal places in the number with the fewest
 4.107 103 S decimal places.
 9.84 105
 0.984 106
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.

Multiplication and Division

In multiplication and division, we are normally limited to the number of digits contained 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  10a means that log n  a (3-1)

1 1
For example, 2 is the logarithm of 100 because 100  102. The logarithm of 0.001 is 3 103    0.001
103 1 000
because 0.001  103. To find the logarithm of a number with your calculator, enter the
number and press the log function.
In Equation 3-1, the number n is said to be the antilogarithm of a. That is, the antilog-
arithm of 2 is 100 because 102  100, and the antilogarithm of 3 is 0.001 because
103  0.001. Your calculator has either a 10 x key or an antilog key. To find the antiloga-
rithm of a number, enter it in your calculator and press 10 x (or antilog).

3-2 Significant Figures in Arithmetic 41

 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  2.530 log 3.39 105  4.470
{{ {{
Characteristic Mantissa Characteristic Mantissa
2  0.530  4  0.470

Number of digits in mantissa of log x  number The number 339 can be written 3.39 102. The number of digits in the mantissa of log
of significant figures in x: 339 should equal the number of significant figures in 339. The logarithm of 339 is properly
log(5.403 108 )  7.267 4 expressed as 2.530. The characteristic, 2, corresponds to the exponent in 3.39 102.
123 123
To see that the third decimal place is the last significant place, consider the following
4 digits 4 digits
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.
In the conversion of a logarithm into its antilogarithm, the number of significant figures
in the antilogarithm should equal the number of digits in the mantissa. Thus,

{  10 {  3.8
{ 10
Number of digits in antilog x (10x )  antilog (3.42) 3.42 4
number of significant figures in mantissa of x:
2 digits 2 digits 2 digits
10 6.142  1.39 106
{ { Here are several examples showing the proper use of significant figures:
3 digits 3 digits
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

Significant Figures and Graphs

Problem 3-8 shows you how to control gridlines When drawing a graph on a computer, consider whether the graph is meant to display quali-
in an Excel graph. tative 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.

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

3-3 Types of Error

Every measurement has some uncertainty, which is called experimental error. Conclusions
can be expressed with a high or a low degree of confidence, but never with complete cer-
tainty. Experimental error is classified as either systematic or random.
Systematic error is a consistent error that can
Systematic Error
be detected and corrected. Box 3-1 describes
Standard Reference Materials designed to Systematic error, also called determinate error, arises from a flaw in equipment or the
reduce systematic errors. design of an experiment. If you conduct the experiment again in exactly the same manner,

42 CHAPTER 3 Experimental Error

the error is reproducible. In principle, systematic error can be discovered and corrected,
although this may not be easy.
For example, a pH meter that has been standardized incorrectly produces a systematic
error. Suppose you think that the pH of the buffer used to standardize the meter is 7.00, but it Ways to detect systematic error:
is really 7.08. Then all your pH readings will be 0.08 pH unit too low. When you read a pH 1. Analyze a known sample, such as a
of 5.60, the actual pH of the sample is 5.68. This systematic error could be discovered by Standard Reference Material. Your method
using a second buffer of known pH to test the meter. should reproduce the known answer. (See
Another systematic error arises from an uncalibrated buret. The manufacturer’s toler- Box 15-1 for an example.)
2. Analyze “blank” samples containing none of
ance for a Class A 50-mL buret is
0.05 mL. When you think you have delivered 29.43 mL,
the analyte being sought. If you observe a
the real volume could be anywhere from 29.38 to 29.48 mL and still be within tolerance. One nonzero result, your method responds to
way to correct for an error of this type is to construct a calibration curve, such as that in Fig- more than you intend. Section 5-1 discusses
ure 3-3, by the procedure on page 38. To do this, deliver distilled water from the buret into a different kinds of blanks.
flask and weigh it. Determine the volume of water from its mass by using Table 2-7. Figure 3. Use different analytical methods to measure
3-3 tells us to apply a correction factor of 0.03 mL to the measured value of 29.43 mL. The the same quantity. If results do not agree,
there is error in one (or more) of the methods.
actual volume delivered is 29.43  0.03  29.40 mL.
4. Round robin experiment: Different people
A key feature of systematic error is that it is reproducible. For the buret just discussed, in several laboratories analyze identical
the error is always 0.03 mL when the buret reading is 29.43 mL. Systematic error may samples by the same or different methods.
always be positive in some regions and always negative in others. With care and cleverness, Disagreement beyond the estimated
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 uncontrolled (and Random error cannot be eliminated, but it
maybe uncontrollable) variables in the measurement. Random error has an equal chance of might be reduced by a better experiment.
being positive or negative. It is always present and cannot be corrected. There is random
error associated with reading a scale. Different people reading the scale in Figure 3-1 report
a range of values representing their subjective interpolation between the markings. One per-
son reading the same instrument several times might report several different readings.
Another random error results from electrical noise in an instrument. Positive and negative
fluctuations occur with approximately equal frequency and cannot be completely eliminated.

Precision and Accuracy

Precision describes the reproducibility of a result. If you measure a quantity several times Precision: reproducibility
and the values agree closely with one another, your measurement is precise. If the values Accuracy: nearness to the “truth”
vary widely, your measurement is not 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.

Box 3-1 Standard Reference Materials Before the introduction of this reference material, five labo-
ratories analyzing identical samples reported a range of results
Inaccurate laboratory measurements can mean wrong medical with relative errors of 40% to 110% of the expected value. After
diagnosis and treatment, lost production time, wasted energy and distribution of the reference material, the error was reduced to
materials, manufacturing rejects, and product liability. The U.S. 20% to 40%.
National Institute of Standards and Technology and national
standards laboratories around the world distribute Standard Refer-
ence 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.1 100 After
Error (% of target value)

For example, in treating patients with epilepsy, physicians

depend on laboratory tests to measure concentrations of anticon- 80
vulsant drugs in blood serum. Drug levels that are too low lead to
60
seizures; high levels are toxic. Because tests of identical serum
specimens at different laboratories were giving an unacceptably 40
wide range of results, the National Institute of Standards and
Technology developed a Standard Reference Material containing 20
known levels of antiepilepsy drugs in serum. The reference mate-
0
rial now enables different laboratories to detect and correct errors Phenytoin Phenobarbital Primidone Ethosuximide
in their assay procedures. (Mysoline) (Zarontin)

3-3 Types of Error 43

 A measurement might be reproducible, but wrong. If you made a mistake preparing a
solution for a titration, you might 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 measure-
ment. The “true” value is best obtained by an experienced person using a well-tested proce-
dure. 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 confidence, but never proof, that results
are accurate.

Absolute and Relative Uncertainty

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

Relative absolute uncertainty

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

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

3-4 Propagation of Uncertainty

from Random Error
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 method. If possi-
ble, uncertainty is expressed as the standard deviation or as a confidence interval, which are
By far, most propagation of uncertainty discussed in Chapter 4. This section applies only to random error. We assume that systematic
computations that you will encounter deal with error has been detected and corrected.
random error, not systematic error. Our goal is For most experiments, we need to perform arithmetic operations on several numbers,
always to eliminate systematic error.
each of which has a random error. The most likely uncertainty in the result is not simply the
sum of the individual errors, because some of them are likely to be positive and some nega-
tive. We expect some cancellation of errors.

Addition and Subtraction

Suppose you wish to perform the following arithmetic, in which the experimental uncertain-
ties, 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?

44 CHAPTER 3 Experimental Error

 For addition and subtraction, the uncertainty in the answer is obtained from the absolute
uncertainties of the individual terms as follows:
Uncertainty in addition
and subtraction: e4  2e21  e22  e23 (3-5) For addition and subtraction, use absolute
uncertainty.
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 express 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 insignifi-
cant 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 significant.
It is sensible to drop the insignificant figures now and express the final result as
3.06 (
0.04) (absolute uncertainty) For addition and subtraction, use absolute
uncertainty. Relative uncertainty can be found
3.06 (
1%) (relative uncertainty)
at the end of the calculation.

Example Uncertainty in a Buret Reading

The volume delivered by a buret is the difference between final and initial readings. 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,
the uncertainty in volume delivered is
0.03 mL.

Multiplication and Division

For multiplication and division, first convert all uncertainties into percent relative uncertain-
ties. Then calculate the error of the product or quotient as follows:
Uncertainty in multiplication
and division: %e4  2 (%e1 ) 2  (%e2 ) 2  (%e3 ) 2 (3-6) For multiplication and division, use 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 into percent relative uncertainties.
1.76 (
1.7 %) 1.89 (
1.1 %)
 5.64
e4 Advice Retain one or more extra insignificant
0.59 (
3.4 %) figures until you have finished your entire
calculation. Then round to the correct number
Then find the percent relative uncertainty of the answer by using Equation 3-6. of digits. When storing intermediate results in a
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 into absolute uncertainty, find 4.0 % of the answer.
4.0 % 5.64  0.040 5.64  0.23

3-4 Propagation of Uncertainty from Random Error 45

 The answer is 5.64 (
0.23 ). Finally, drop the insignificant digits.
For multiplication and division, use percent 5.6 (
0.2) (absolute uncertainty)
relative uncertainty. Absolute uncertainty can
5.6 (
4% ) (relative uncertainty)
be found at the end of the calculation.
The denominator of the original problem, 0.59, limits the answer to two digits.

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 into 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)
The result of a calculation ought to be written Because the uncertainty begins in the 0.01 decimal place, it is reasonable to round the result
in a manner consistent with its uncertainty. to the 0.01 decimal place:
0.62 (
0.02) (absolute uncertainty)
0.62 (
3% ) (relative uncertainty)

The Real Rule for Significant Figures

The real rule: The first uncertain figure is the last The first digit of the absolute uncertainty is the last significant digit in the answer. For
significant figure. example, in the quotient
0.002 364 (
0.000 003)
 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 0.094 6
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.
It is all right to keep one extra digit when an Now you can appreciate why it is all right to keep one extra digit when an answer lies
answer lies between 1 and 2. between 1 and 2. The quotient 82/80 is better written as 1.02 than 1.0. If I write 1.0, you can
surmise that the uncertainty is at least 1.0
0.1 
10%. The actual uncertainty lies in the
second decimal place, not the first decimal place.

Example Significant Figures in Laboratory Work

You prepared a 0.250 M NH3 solution by diluting 8.45 (
0.04) mL of 28.0 (
0.5) wt%
NH3 [density  0.899 (
0.003) g/mL] up to 500.0 (
0.2) mL. Find the uncertainty in
0.250 M. The molecular mass of NH3, 17.030 5 g/mol, has negligible uncertainty relative
to other uncertainties in this problem.

46 CHAPTER 3 Experimental Error

 Solution To find the uncertainty in molarity, we need to find the uncertainty in moles The rationale for finding the uncertainty in
delivered to the 500-mL flask. The concentrated reagent contains 0.899 (
0.003) g of the molecular mass of NH3 is explained in
solution per milliliter. Weight percent tells us that the reagent contains 0.280 (
0.005) g Section 3-5:

of NH3 per gram of solution. In our calculations, we retain extra insignificant digits and N: 14.006 7
0.000 2
round off only at the end. 3H: 3(1.007 94
0.000 07)

Grams of NH3 per g solution g NH3 N: 14.006 7

0.000 2
mL in concentrated  0.899 (
0.003) mL
0.280 (
0.005)
g solution 3H: 3.023 82
0.000 21
reagent NH3: 17.030 52
20.000 22  0.000 212
g solution g NH3  17.030 52
0.000 29
 0.899 (
0.334% ) 0.280 (
1.79%)  17.030 5
0.000 3
mL g solution
g NH3
 0.251 7 (
1.82%)
mL
because 2 (0.334%) 2  (1.79%) 2  1.82%. Convert absolute uncertainty into percent
Next, we find the moles of ammonia contained in 8.45 (
0.04) mL of concentrated relative uncertainty for multiplication.
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 5 (
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

L 0.500 0 (
0.04%) L
 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  x a, 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 y to the
Uncertainty for 0.333 333 333 . . . power with the y x button.
powers and roots: y  x a 1 %ey  a(%ex ) (3-7)

For example, if y  1x  x12, a 2% uncertainty in x will yield a 1122 (2%)  1% uncertainty

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

The natural logarithm (ln) of x is the number y, whose value is such that x  e y, 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
natural logarithm: y  ln x 1 ey  (3-9)
x
Now consider y  antilog x, which is the same as saying y  10 x. In this case, the rela-
tive uncertainty in y is proportional to the absolute uncertainty in x.

3-4 Propagation of Uncertainty from Random Error 47

 Box 3-2 Propagation of Uncertainty in the Product x
x

Table 3-1 says that the uncertainty in the function y  x a is %ey  majority of cases, the uncertainty in the product x
z is not as
a(%ex ). If y  x2, then a  2 and %ey  2(%ex ). A 3% uncer- great as the uncertainty in x2.
tainty in x leads to a (2)(3%)  6% uncertainty in y. Example. The distance traveled by a falling object in time
But what if we just apply the multiplication formula 3-6 to t is 12gt 2, where g is the acceleration of gravity. If t has an uncer-
the product x
x? tainty of 1%, the uncertainty in t2 is 2(%et )  2(1% )  2%. The
uncertainty in distance computed from 12gt2 will also be 2%. If you
x(
e1 )
x(
e2 )  x2 (
e3 )
(incorrectly) used Equation 3-6, you would compute an uncer-
%e3  2 (%e1 ) 2  (%e2 ) 2  2 (3%) 2  (3%) 2  4.2 % tainty in distance of 21% 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 com-
puted 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.
Equation 3-6 presumes that the uncertainty in each factor of
the product x
z is random and independent of the other. In the
Relative Relative Relative Relative
product x
z, the measured value of x could be high sometimes uncertainty uncertainty uncertainty uncertainty
and the measured value of z could be low sometimes. In the in x in z in x · z in x 2

Uncertainty ey
y  10 x 1  (ln 10)ex ⬇ 2.302 6 ex (3-10)
for 10 x: y
If y  ex, the relative uncertainty in y equals the absolute uncertainty in x.

Uncertainty ey
y  ex 1  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 rules
propagation of random uncertainty for any for exponents, logs, and antilogs, but you should be able to use them.
function.

Example Uncertainty in Hⴙ 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
10 log [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  10 x, in which y  [H] and x  (5.21
0.03).
For y  10 x, 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 1 1 ey  4.26 107
y 6.17 106
The concentration of H is 6.17 (
0.426) 106  6.2 (
0.4) l06 M. An
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.

48 CHAPTER 3 Experimental Error

Table 3-1 Summary of rules for propagation of uncertainty
Function Uncertainty Functiona Uncertaintyb
y  x1  x2 ey  2e2x1  e2x2 y  xa %ey  a%ex
1 ex ex
y  x1  x2 ey  2e2x1  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
a . x represents a variable and a represents a constant that has no uncertainty,
b. ex /x is the relative error in x and %ex is 100 ex /x.

3-5 Propagation of Uncertainty:

Systematic Error
Systematic error occurs in some common situations and is treated differently from random
error in arithmetic operations.

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. The uncertainty is not
mainly from random error in measuring the atomic mass. The uncertainty is predominantly
from isotopic variation in samples of oxygen from different sources. That is, oxygen from
one source could have a mean atomic mass of 15.999 1 and oxygen from another source
could have an atomic mass of 15.999 7. The atomic mass of oxygen in a particular lot of
reagent has a systematic uncertainty. It could be relatively constant at 15.999 7 or 15.999 1,
or any value in between, with only a small random variation around the mean value.
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 of the mass of n
atoms is n (uncertainty of one atom)  2 (
0.000 3) 
0.000 6. The uncertainty is
not
20.000 32  0.000 32 
0.000 42. For systematic uncertainty, we add the uncertain-
ties of each term in a sum or difference.
Now let’s 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 Propagation of systematic uncertainty:
4H: 4(1.007 94
0.000 07)  4.031 76
0.000 28 d 4 0.000 07 Uncertainty in mass of n identical atoms 
n (uncertainty in atomic mass).
28.053 16
? (3-13)
For the uncertainty in the sum of the masses of 2C  4H, we use Equation 3-5, which
applies to random error, 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 We choose to use the rule for propagation of
random uncertainty for the sum of atomic
28.053 16
0.001 6
masses of different elements.
28.053
0.002 g/mol

Multiple Deliveries from One Pipet

A 25-mL Class A volumetric pipet is certified by the manufacturer to deliver
25.00
0.03 mL. The volume delivered by a given pipet is reproducible, but can be any-
where in the range 24.97 to 25.03 mL. The difference between 25.00 mL and the actual vol-
ume delivered by a particular pipet is a systematic error. It is always the same, within a small
random error. You could calibrate a pipet by weighing the water it delivers, as in Section 2-9.

3-5 Propagation of Uncertainty: Systematic Error 49

 Calibration eliminates systematic error, because we would know that the pipet always deliv-
ers, say, 25.991
0.006 mL. The remaining uncertainty (
0.006 mL) is random error.
If you use an uncalibrated 25-mL Class A volumetric pipet four times to deliver a total of
100 mL, what is the uncertainty in 100 mL? Because the uncertainty is a systematic error, the
uncertainty in four pipet volumes is like the uncertainty in the mass of 4 moles of oxy-
gen: The uncertainty is
4 0.03 
0.12 mL, not
20.032  0.032  0.032  0.032 

0.06 mL.
Calibration improves accuracy. Suppose that a calibrated pipet delivers a mean volume
of 24.991 mL with a standard deviation (a random variation) of
0.006 mL. If you deliver
In this example, calibration reduces the four aliquots from this pipet, the volume delivered is 4 24.991  99.964 mL and the
uncertainty from
0.12 mL to
0.012 mL. uncertainty is
20.0062  0.0062  0.0062  0.0062 
0.012 mL.

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

Summary
The number of significant digits in a number is the minimum value). Systematic error can be discovered and eliminated by a
required to write the number in scientific notation. The first uncer- clever person, but some random error is always present. For ran-
tain digit is the last significant figure. In addition and subtraction, dom errors, propagation of uncertainty in addition and subtrac-
the last significant figure is determined by the number with the tion requires absolute uncertainties (e3  2e21  e22 ), whereas
fewest decimal places (when all exponents are equal). In multiplica- multiplication and division utilize relative uncertainties
tion and division, the number of figures is usually limited by the fac- (%e3  2 %e21  e22 ). Other rules for propagation of random error
tor with the smallest number of digits. The number of figures in the are found in Table 3-1. Always retain more digits than necessary
mantissa of the logarithm of a quantity should equal the number of during a calculation and round off to the appropriate number of dig-
significant figures in the quantity. Random (indeterminate) error its at the end. Systematic error in atomic mass or the volume of a
affects the precision (reproducibility) of a result, whereas systematic pipet leads to larger uncertainty than we get from random error. We
(determinate) error affects the accuracy (nearness to the “true” always strive to eliminate systematic errors.

Exercises
3-A. Write each answer with a reasonable number of figures. Find 3-C. We have a 37.0 (
0.5) wt% HCl solution with a density of
the absolute and percent relative uncertainty for each answer. 1.18 (
0.01) g/mL. To deliver 0.050 0 mol of HCl requires 4.18 mL
(a) [12.41 (
0.09)  4.16 (
0.01)] 7.068 2 (
0.000 4)  ? of solution. If the uncertainty that can be tolerated in 0.050 0 mol is
(b) [3.26 (
0.10) 8.47 (
0.05)]  0.18 (
0.06)  ?
2%, how big can the absolute uncertainty in 4.18 mL be?
(c) 6.843 (
0.008) 104  [2.09 (
0.04)  1.63 (
0.01) ]  ? (Caution: In this problem, you have to work backward. You would
(d) 23.24
0.08  ? normally compute the uncertainty in mol HCl from the uncertainty
(e) (3.24
0.08) 4  ? in volume:
(f) log(3.24
0.08)  ? g solution g HCl
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 53.4 wt% mol HCl
NaOH will you need to prepare 2.000 L of 0.169 M NaOH? But, in this case, we know the uncertainty in mol HCl (2%) and we
(b) If the uncertainty in delivering NaOH is
0.01 mL, calculate need to find what uncertainty in mL solution leads to that 2% uncer-
the absolute uncertainty in the molarity (0.169 M). Assume there is tainty. The arithmetic has the form a  b c d, for which %e2a 
negligible uncertainty in the formula mass of NaOH and in the final %e2b  %e2c  %e2d. If we know %ea, %ec, and %ed, we can find
volume (2.000 L). %eb by subtraction: %e2b  %e2a  %e2c  %e2d.)

Problems
Significant Figures (c) 0.135 2 to 3 significant figures
3-1. How many significant figures are there in the following (d) 2.051 to 2 significant figures
numbers? (e) 2.005 0 to 3 significant figures
(a) 1.903 0 (b) 0.039 10 (c) 1.40 104 3-3. Round each number to three significant figures:
3-2. Round each number as indicated: (a) 0.216 74
(a) 1.236 7 to 4 significant figures (b) 0.216 5
(b) 1.238 4 to 4 significant figures (c) 0.216 500 3

50 CHAPTER 3 Experimental Error

 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.

3-4. Vernier scale. The figure above shows a scale found on instru- overrule you several times. Continue to reset the limits as you want
ments such as a micrometer caliper used for accurately measuring them and click OK each time until the graph looks the way you
dimensions of objects. The lower scale slides along the upper scale intend. To add gridlines, click in the graph, go to the CHART menu
and is used to interpolate between the markings on the upper scale. and select CHART OPTIONS. Select the Gridlines tab and check both
In (a), the reading (at the left-hand 0 of the lower scale) is between sets of Major gridlines and Minor gridlines and click OK. In the
1.4 and 1.5 on the upper scale. To find the exact reading, observe CHART OPTIONS menu, select the Legend tab and deselect Show
which mark on the lower scale is aligned with a mark on the upper legend. Move the x-axis numbers from the middle of the chart to the
scale. Because the 6 on the lower scale is aligned with the upper bottom as follows: Double click the y-axis (not the x-axis) and
scale, the correct reading is 1.46. Write the correct readings in select the Scale tab. Set “Value (x) axis crosses at” to 0.04. Click
(b) and (c) and indicate how many significant figures are in each OK and the volume labels move beneath the graph. Your graph
reading. should look the same as Figure 3-3.
3-5. Write each answer with the correct number of digits.
(a) 1.021  2.69  3.711 Volume (mL) Correction (mL)
(b) 12.3  1.63  10.67 0.03 0.00
(c) 4.34 9.2  39.928 10.04 0.04
(d) 0.060 2  (2.113 104 )  2.849 03 106 20.03 0.02
(e) log(4.218 1012 )  ? 29.98 0.03
(f) antilog(3.22)  ? 40.00 0.00
(g) 102.384  ? 49.97 0.03
3-6. Write the formula mass of (a) BaF2 and (b) C6H4O4 with a rea-
sonable number of digits. Use the periodic table inside the cover of Types of Error
this book to find atomic masses. 3-9. Why do we use quotation marks around the word true in the
3-7. Write each answer with the correct number of significant figures. statement that accuracy refers to how close a measured value is to
(a) 1.0  2.1  3.4  5.8  12.300 0 the “true” value?
(b) 106.9  31.4  75.500 0 3-10. Explain the difference between systematic and random errors.
(c) 107.868  (2.113 102 )  (5.623 103 )  5 519.568 3-11. Suppose that in a gravimetric analysis, you forget to dry the
(d) (26.14/37.62) 4.38  3.043 413 filter crucibles before collecting precipitate. After filtering the prod-
(e) (26.14/ (37.62 108 )) (4.38 102 )  3.043 413 1010 uct, you dry the product and crucible thoroughly before weighing
(f) (26.14/3.38)  4.2  11.933 7 them. Is the mass of product always high or always low? Is the
(g) log(3.98 104 )  4.599 9 error in mass systematic or random?
(h) 106.31  4.897 79 107
3-12. State whether the errors in parts (a)–(d) are random or
3-8. Controlling the appearance of a graph in Excel. systematic:
Figure 3-3 requires gridlines to read the graph for buret corrections. (a) A 25-mL transfer pipet consistently delivers 25.031
0.009 mL.
The purpose of this exercise is to format a graph so it looks like (b) A 10-mL buret consistently delivers 1.98
0.01 mL when
Figure 3-3. Follow the procedure in Section 2-11 to make a graph of drained from exactly 0 to exactly 2 mL and consistently delivers
the data in the following table. The Chart Type is xy Scatter show- 2.03 mL
0.02 mL when drained from 2 to 4 mL.
ing data points connected by straight lines. Double click on the (c) A 10-mL buret delivered 1.983 9 g of water when drained from
x-axis and select the Scale tab. Set Minimum  0, Maximum  50, exactly 0.00 to 2.00 mL. The next time I delivered water from the
Major unit  10, and Minor unit  1. Select the Number tab and 0.00- to the 2.00-mL mark, the delivered mass was 1.990 0 g.
highlight Number. Set Decimal places  0. In a similar manner, set (d) Four consecutive 20.0- L injections of a solution into a chro-
the ordinate to run from 0.04 to 0.05 with a major unit of 0.02 matograph were made and the area of a particular peak was 4 383,
and a minor unit of 0.01, as in Figure 3-3. The spreadsheet may 4 410, 4 401, and 4 390 units.

Problems 51
 Cheryl Cynthia Carmen Chastity

Figure for Problem 3-13.

3-13. Cheryl, Cynthia, Carmen, and Chastity shot the targets above (b) To prepare a solution of NaCl, you weigh out 2.634 (
0.002) g
at Girl Scout camp. Match each target with the proper description. and dissolve it in a volumetric flask whose volume is
(a) accurate and precise 100.00 (
0.08) mL. Express the molarity of the solution, along
(b) accurate but not precise with its uncertainty, with an appropriate number of digits.
(c) precise but not accurate 3-20. What is the true mass of water weighed at 24°C in the air if
(d) neither precise nor accurate the apparent mass is 1.034 6
0.000 2 g? The density of air is
3-14. Rewrite the number 3.123 56 (
0.167 89%) in the forms 0.001 2
0.000 1 g/mL and the density of balance weights is
(a) number (
absolute uncertainty) and (b) number (
percent rel- 8.0
0.5 g/mL. The uncertainty in the density of water in Table 2-7
ative uncertainty) with an appropriate number of digits. is negligible in comparison to the uncertainty in the density of air.
Propagation of Uncertainty 3-21. Twelve dietary iron tablets were analyzed by the gravimetric
procedure in Section 1-4 and the final mass of Fe2O3 (FM 159.688)
3-15. Find the absolute and percent relative uncertainty and express
was 0.2774
0.0018 g. Find the average mass of Fe per tablet.
each answer with a reasonable number of significant figures.
(Relative uncertainties in atomic masses are small compared with
(a) 6.2 (
0.2)  4.1 (
0.1)  ?
relative uncertainty in the mass of Fe2O3. Neglect uncertainties in
(b) 9.43 (
0.05) 0.016 (
0.001)  ?
atomic masses in this problem.)
(c) [6.2 (
0.2)  4.1 (
0.1) ]  9.43 (
0.05)  ?
(d) 9.43 (
0.05) 5[6.2 (
0.2) 103 ]  [4.1 (
0.1) 103 ]6  ? 3-22. We can measure the concentration of HCl solution (a proce-
dure called standardizing the solution) by reaction with pure
3-16. Find the absolute and percent relative uncertainty and express
sodium carbonate: 2H  Na2CO3 S 2Na  H2O  CO2. A vol-
each answer with a reasonable number of significant figures.
ume of 27.35
0.04 mL of HCl solution was required for
(a) 9.23 (
0.03)  4.21 (
0.02)  3.26 (
0.06)  ?
complete reaction with 0.967 4
0.000 9 g of Na2CO3
(b) 91.3 (
1.0) 40.3 (
0.2)/21.1 (
0.2)  ?
(FM 105.988
0.001). Find the molarity of the HCl and its
(c) [4.97 (
0.05)  1.86 (
0.01) ] /21.1 (
0.2)  ?
absolute uncertainty.
(d) 2.016 4 (
0.000 8)  1.233 (
0.002)  4.61 (
0.01)  ?
(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  ? measured properties of pure crystalline silicon:2 (1) atomic mass
(f) [3.14 (
0.05) ] 1/3  ? (obtained from the mass and abundance of each isotope), (2) den-
(g) log[3.14 (
0.05)]  ? sity of the crystal, (3) size of the unit cell (the smallest repeating
unit in the crystal), and (4) number of atoms in the unit cell. For sil-
3-17. Verify the following calculations:
icon, the mass is mSi  28.085 384 2 (35) g/mol, where 35 is the
(a) 23.141 5 (
0.001 1)  1.772 43 (
0.000 31 )
standard deviation in the last two digits. The density is ρ 
(b) log[3.141 5 (
0.001 1)]  0.497 14 (
0.000 15 )
2.329 031 9 (18) g/cm3, the size of the cubic unit cell is c0 
(c) antilog[3.141 5 (
0.001 1)]  1.3852 (
0.0035 ) 103
5.431 020 36 (33) 108 cm, and there are 8 atoms per unit cell.
(d) ln[3.141 5 (
0.001 1)]  1.144 70 (
0.000 35 )
Avogadro’s number is computed from the equation
b  0.800 (
0.015 )
20.104 (
0.006)
(e) loga mSi
0.051 1 (
0.000 9)
NA 
3-18. Express the molecular mass (
uncertainty) of C9H9O6N3 with (ρc 30 )/8
the correct number of significant figures. From the measured properties and their uncertainties (standard
3-19. (a) Show that the formula mass of NaCl is 58.443 deviations), compute Avogadro’s number and its uncertainty. To
(
0.002) g/mol. find the uncertainty of c 30, use the function y  x a in Table 3-1.

52 CHAPTER 3 Experimental Error