Factors Influencing the Creep of Dental Amalgam D.B. MAHLER and J.D. ADEY Oregon Health Sciences University, 611 SW Campus Drive, Portland, Oregon 97201 The purpose of this study was to investigate the influence of relevant microstructural and compositional factors on the creep of 12 representative dental amalgams by means of stepwise multiple linear regression. The independent variables accepted by the regression were volume percent ofthe p (Cu6Sn5) phase, grain size of the yl (Ag2Hg3) phase, volume percent of the y (Ag3Sn) plus E (Cu3Sn) phases, number of very small p crystals (< 1.5pm)per mm, and weight percent of Hg. The results of this regression showed an adjusted R2 of 0.949, significant at p = 0.002. J Dent Res 70(11):1394-1400, November, 1991
Introduction. The significance of creep as a factor relating to the clinical performance of dental amalgams was initially shown over 20 years ago (Mahler et al., 1970) and has since been demonstrated repeatedly. Because ofthis relationship, creep is considered ofmajor importance in characterizing dental amalgam and forms a significant part ofthe American Dental Association Specification No. 1. Information relevant to this physical property would be ofvalue in understanding the basic metallurgical and mechanical behavior of amalgam, as well as the relationship of this behavior to clinical performance. Because this paper will frequently refer to the various metallic phases of dental amalgam, these phases and their descriptions appear in Table 1. By measuring the creep ofthe y, y 1, and Y2 phases separately, Espevik (1977) concluded that the creep of amalgam is primarily due to creep of the y I matrix phase. Factors that have been shown to influence creep include final Hg content of the set amalgam (Mahler and Van Eysden, 1969), Sn in the y I phase (Mahler and Adey, 1979), Zn in the y1 phase (Johnson and Paffenbarger, 1980), grain size of y 1 (Mahleret al., 1977), amount ofthe Y2 phase (Mahler et al., 1977; Osborne and Gale, 1990), and the amount ofthe y phase (Mahler et al., 1977). In addition, the presence of small pe crystals in the y 1 matrix has been suggested as having an influence on creep (Okabe et al., 1978). Other factors that could prove to be relevant are the volume percent of n', because the amount of this phase may be related to the completeness of elimination of y2, and the total volume percent of all of the hard particles that would inhibit the creep of y 1, i.e., y + E + E + p'. In previous studies, the influence of some of these factors has been demonstrated individually or for individual alloys under various conditions. Because of this selective approach, it is not known which factors bear a relationship to each other or which factors are most significant in influencing creep. The purpose of this study was to measure all of these factors on a group of representative dental amalgams that exhibit a wide range of creep values. This would be followed by a correlation analysis and a stepwise multiple linear regression analysis, for determination of the relationships among factors and the factors that are most significant in influencing creep.
Received for publication February 1, 1991 Accepted for publication May 23, 1991 This investigation was supported by USPHS Research Grant DE 02936 from the National Institute of Dental Research, National Institutes of Health, Bethesda MD 20892.
1394
TABLE 1 GREEK LETTER DESIGNATION FOR AMALGAM PHASES y
Ag-Sn phase in the amalgam alloy (Ag3Sn, but may also include the B Ag-Sn phase)
Y1
Ag-Hg reaction matrix phase (Ag22SnHg27)
Y2
Sn-Hg reaction phase (Sn8Hg)
E
Cu-Sn phase in the amalgam alloy (Cu3Sn)
A'
Cu-Sn reaction phase (Cu6Sn5)
E Ag-Cu eutectic (approximate) Note: The formulations expressed in parentheses demonstrate the major elements present in these phases and their approximate compositions.
Materials and methods. Twelve amalgam alloys were investigated and are listed in Table 2 together with their mixing conditions. Because ofthe influence of age and batch variation, mixing conditions were determined by subjective tests for adequate trituration and mix plasticity and may not conform to the manufacturers' instructions. The volume percent of Y2 and the creep values ofthe amalgams made from these alloys, as determined in this study, are also listed in Table 2. The methods used to determine these two characteristics will be described subsequently. These alloys were selected to include different types of amalgams, as indicated by the volume percent of Y2, as well as those having different creep values. In Table 2, five amalgams can be classified as low-copper y 2-containing, four amalgams as highcopper y2-free, and three amalgams as high-copper containing small amounts of Y2 ranging from 0.7 to 3.6 volume percent. The particles for the 12 amalgam alloys were lathe-cut for four alloys, spherical for four alloys, and admixtures of lathe-cut and spherical for four alloys. Thus, the materials selected for study were reasonably representative of dental amalgam alloys, and their amalgams demonstrated a wide range of creep values. In the stepwise regression, creep was the dependent variable, and the factors investigated for influence on creep were the independent variables (IV's). The factors investigated were selected on the basis of their physical relevance to the mechanism of creep, as well as the results of the studies described in the "Introduction". In Table 4, nine selected factors are listed with their respective code designations, which will be used in the "Results" and "Discussion" sections. Also listed in this Table are the individual correlation coefficients of each of these factors to creep and their corresponding significance levels. Creep was determined in accordance with the American Dental Association Specification No. 1 for Alloy for Dental Amalgam. However, to ensure more stability in the phase structure, two weeks of storage at 370C was used, rather than the one week of the Specification, and the tests for all factors were conducted on.twoweek-old specimens. For creep, five replicate tests were run and averaged. Because the creep values ranged over three orders of
Downloaded from jdr.sagepub.com at UNIV OF MICHIGAN on June 18, 2015 For personal use only. No other uses without permission.
CREEP OF AMALGAM
Vol. 70 No. 11
1395
TABLE 2
ALLOYS, MIXING CONDITIONS, TYPE, AND CREEP
Mixing Conditions
Type
Code
Alloy Batch No.
Manufacturer
%Hg
Trituration*
y2 (vol%)
Creep (%)
AR
Aristaloy 0511781
Baker Dental, Carteret, NJ
55.0
M10
9.9
3.765
OP
Optaloy 30F70
L.D. Caulk, Milford, DE
54.0
M9
9.9
3.369
II
Optaloy II 060176
L.D. Caulk
52.5
M9
3.6
1.772
PH
Phasealloy 423/062983
Phasealloy, El Cajon, CA
51.5
M13
1.2
1.610
NT
N.T. Dentalloy 120475
S.S. White, Philadelphia, PA
49.5
M9
10.7
1.532
SS
Spherical Alloy 297501
Shofu Dental, Menlo Park, CA
44.5
M7
10.0
1.207
NZ
Dispersalloy, non-Zn 8042 J & J Ltd, Montreal, Canada
52.5
M13
0.7
1.014
DS
Dispersalloy 31426
J & J Dent, E. Windsor, NJ
51.5
M12
0.0
0.674
SP
Spheraloy T155
Kerr Mfg, Romulus, MI
47.0
M11**
9.8
0.523
80
Epoque 80 7-217/82.03
Scania Dental, Knivsta, Sweden
52.5
M7
0.0
0.222
SY
Sybraloy 1163
Kerr Mfg
48.0
H9
0.0
0.095
M9** 0.083 0.0 44.0 S.S. White TY Tytin 2038403 * Alloys mixed in a re-usable capsule with pestle (J & J Dent. Prod.) using a Varimix III amalgamator (L.D. Caulk Co.). Amalgamator speed setting and time in seconds are indicated. ** Alloys mixed as above but without a pestle.
magnitude, the natural log of 100 times the creep average was used as the dependent variable in the regression. Specimen preparation and the grinding and polishing procedure for SEM (Model JSM-T330A, JEOL USA Inc., Peabody, MA) and microprobe analysis (Model EMX-SM, Applied Research Laboratories, Sunland, CA) consisted of the following: Specimens measuring 4 mm (diam.) x 8 mm (length) were made according to ADA Specification No. 1. The specimens were mounted in brass rings with conductive epoxy resin and ground and polished in the following sequence: silicon carbide papers, 320-, 400-, and 600-grit; diamond paste, 15-, 9-, 3-, 1-, and 1/4-pum particle size. The grain size of y1 was measured on SEM photographs of etched amalgams at magnifications of 5000 x and 10,000 x, depending on the sizes of the grains. Etching was done with solution No. 1 of Wings etch (Wing, 1965) at 1/10 strength in a Minimet polisher on a microcloth (Buehler Ltd., Lake Bluff, IL), with a pressure setting of zero. A sheet of acetate paper with evenly spaced lines was placed over the SEM photographs, and the numbers of grains lying beneath fixed line lengths were counted and converted to an average grain size in pm. From 50 to 90 line lengths were utilized to produce an overall average grain size. Although this method underestimates the true grain size, the difference is the same for all alloys and would not be of consequence when used in a regression analysis. Sn in the y 1 phase was measured using the following microprobe settings: excitation potential = 15 keV; sample current = 5 nA; counting time = 60 s; analyzed lines - Ag La, Hg Mcx, Sn Lc, and Cu Kca. For determination of weight percent, the raw counts were corrected for atomic number, absorption, and fluorescence effects (ZAF) by use of the program by Colby (1968) known as MAGIC IV. Zn content ofthe amalgam alloy was measured by flame atomic absorption spectrophotometry(Model IL 151, Instrumentation Laboratories, Inc., Wilmington, MA). The accuracy of measurement was + 0.0 1% Zn, and the average ofthree replicate specimens was used.
The Hg content of the set amalgam was determined by the method of Crawford and Larson (1955). The five specimens tested for creep were used for determination of the average weight percent of Hg. All of the variables indicating the amount of each phase in the amalgam microstructure were measured by x-ray line profiles using lineal analysis (DeHoff and Rhines, 1968). The microprobe beam was kept stationary, and the specimen moved under the beam at a rate of 1.3 pm/s. A four-pen recorder registered the simultaneous signals of the Ag Lcx, Hg Mcx, Sn La, and Cu KaL lines. The individual phases were identified by the appropriate line peaks, and the width of the best element line for each phase was measured at 1/2 the maximum peak height. For determination of the volume percent ofeach phase, all ofthe calculated widths were summed and divided by the total distance traveled on the specimen. A distance of from 4 to 7 mm was traced on each specimen. Variances for this lineal analysis were determined by the method described by Hilliard and Cahn (1961). For determination of the number of small q' crystals per mm, Cu peaks whose widths at 1/2 their peak heights corresponded to less than 1.5 pm in actual size were counted and divided by the total distance traversed. There was no variance associated with this measurement. Performance of stepwise regression is considered tenuous when dealing with a large number of independent variables and a relatively small sample size (Cohen and Cohen, 1983). Therefore, the following procedure was used to reduce the number of independent variables to be considered in the stepwise regression: (1) A correlation analysis was performed among the independent variables (IVs). (2) From this analysis, the IV's were separated into two groups. Group A included the IV's which were not highly correlated with any other IV. Group B included the IVs which were all highly correlated with each other (Table 5). (3) Separate, stepwise, multiple linear regressions were then
Downloaded from jdr.sagepub.com at UNIV OF MICHIGAN on June 18, 2015 For personal use only. No other uses without permission.
J Dent Res November 1991
MAHLER & ADEY
1396
0
R2= 0.949
w
r=
a-
-0.845
U,
W
w U
a-
44
w LL
0 0
x0
z -i
0
3
*
z
+
LOW Cu
-I
10%Y2
® HIGH Cu I-44%Y2
n1
2 0
HIGH Cu
LOW Cu
10%a2
- 4% 2 HIGH Cu 0%0'2
(3 HIGH Cu
W
°%'Y2
I
I
5
10
I 15
I
20
25 LN 100
7'- VOLUME %
Fig. 1-Measured creep vs. volume percent of p'.
x
CREEP
-
CALCULATED
Fig. 2-Measured creep vs. creep calculated from the regression equation.
in Table 5. Based on these coefficients, the independent variables separated into the two groups, A and B, shown in this Table. With two exceptions, Group A lists the IV's not highly correlated with any other IV (p 0.05), while Group B lists the IVs that were highly correlated to each other (p < 0.001). In Table 6, the results of the trial stepwise multiple linear regressions are presented. The column labeled "Starting Matrix" shows the variables used for initiation of the regression procedure. Results. Group A variables by themselves were run first. Subsequently, all Tables 3A and 3B list the data for the 12 alloys and nine factors of the five 's of Group A were used together with each of the IV's evaluated in order of decreasing values for creep. In Table 3A, Sn in of Group B in separate regressions. The column labeled "Variables Accepted" are those IV's of the the y 1, phase was relatively high for the y 2-containing amalgams and decreased significantly as Y2 was eliminated. The grain size of starting matrix that were accepted by the stepwise regression with yI showed significant differences, particularly for the relatively an overall a set for 0.15 (Bendel and Afifi, 1977). The column large grains of alloy SP and alloy TY. The Hg contents ofthe various labeled "Adjusted Re" indicates the R2 values for the multiple linear amalgams as condensed by the method of ADA Specification No. 1 regressions of the IV's listed in the "Variables Accepted" column. are in some cases extremely low compared with what would be The term "Adjusted" means that the raw R2 values have been expected under clinically simulated condensation procedures adjusted for the number of IV's in the equation and the number of cases. This parameter is a more conservative estimate ofthe percent (Mahler, 1979). The weight percent of Zn revealed four alloys that were essen- of variance explained, especially when the sample size is small tially free of Zn (SS, NZ, SY, and TY), while the remaining eight (Cohen and Cohen, 1983). The highest RI of0.949 was achieved by use ofthe five variables: alloys showed varied Zn contents up to approximately 1%. In Table 3B, the volume percent of p was zero for the amalgams volume percent of p', the grain size of y the volume percent of y containing approximately 10% y2 and increased to values up to 24% + E, the number of small crystalsper mm, and the weight percent for the y 2-free amalgams. The volume percent of Y2 was essentially of Hg in the set amalgam. The results of this regression are detailed 10% in the five low-Cu alloys (AR, OP, NT, SS, and SP), 1-4% in three in Table 7, where the listed order of the five variables reflects the dispersed-phase alloys (II, PH, and NZ), and 0% in one dispersed- relative influence of these variables on increasing R2, i.e., their phase alloy (DS) and in the three high-Cu single- composition alloys relative ability to reduce the average variance ofthe points around the regression line. Thus, p' volume percent was the most influen(80, SY, and TY). Table 4 lists the correlation coefficients between creep and each tial factor affecting creep, y grain size was the next most influential ofthe IV's. The first four variables listed showed significant correla- factor, and so on down the column labeled "Ln Creep vs". Changes tions with creep at p < 0.01. These were the volume percent of the 9' in the adjusted R12 and the Student t and p values for each factor are phase, the total volume percent ofy + E + E + n', the weight percent presented as each IV is added to the equation. The t and corresponding ofSn in the y 1 phase, and the number of small crystalsper mm of the p values associated with each factor indicate the significance of the 9' phase. Of these factors, the one producing the highest correlation contribution of that factor to the best-fitting regression. All factors with creep proved to be the volume percent of the p' phase with a had t values significant at p < 0.05, except for the weight percent of correlation coefficient of-0.845 (significant at p < 0.001). A plot of In Hg, for which p = 0.055, slightly higher than 0.05. However, this (creep) vs. volume percent of n 'is shown in Fig. 1. [Vwas retained because without its presence, one of the test alloys Significant correlations among the IV's themselves can be seen was found to be an outlier. performed on the IV's of Group A together with each of the IV's of Group B. (4) The most significant factors influencing creep were identified as those producing a regression having the highest R12. The stepwise multiple linear regression was run by the computer program known as SYSTAT (Systat, Inc., Evanston, IL).
were
1,
1
Downloaded from jdr.sagepub.com at UNIV OF MICHIGAN on June 18, 2015 For personal use only. No other uses without permission.
1397
CREEP OF AMALGAM
Vol. 70 No. 11
TABLE 3A DATA USED IN THE REGRESSION Creep (%) n=5
Sn in y1 (Wt%) n=8
Y 1 Grain Size
Hg Content
(pm) n=75
(wt%) n=5
Zn in y (wt%) n=3
AR
3.77 (0.193)
2.79 (0.307)
1.13 (0.352)
46.7 (0.62)
0.53 (0.022)
OP
3.37 (0.460)
2.94 (0.281)
1.14 (0.449)
45.7 (0.29)
0.31 (0.009)
II
1.77 (0.380)
2.32 (0.182)
1.06 (0.426)
45.4 (0.72)
0.28 (0.017)
PH
1.61 (0.075)
1.55 (0.324)
1.32 (0.339)
44.6 (0.14)
0.24 (0.011)
NT
1.53 (0.410)
2.58 (0.145)
1.01 (0.266)
42.1 (0.37)
0.96 (0.003)
SS
1.21 (0.306)
2.46 (0.170)
1.53 (0.701)
42.9 (0.26)
0
NZ
1.01 (0.122)
2.12 (0.244)
1.53 (0.549)
46.7 (0.36)
0
DS
0.67 (0.070)
1.48 (0.345)
1.65 (1.556)
44.1 (0.25)
1.02 (0.020)
SP
0.52 (0.233)
2.37 (0.321)
2.02 (0.831)
44.2 (0.24)
0.50 (0.024)
80
0.22 (0.091)
0.98 (0.359)
1.39 (0.429)
44.5 (0.07)
0.10 (0.025)
SY
0.10 (0.038)
0.69 (0.285)
1.28 (0.499)
43.1 (0.56)
0.02 (0.013)
TY
0.08 (0.045)
1.52 (0.341)
3.49 (2.878)
40.7 (0.16)
0
Alloy
Values are means and standard deviations. TABLE 3B DATA USED IN THE REGRESSION -yEfl'
Alloy AR
OP
(vol%) - 0 -
0
(vol%) 36.1 (2.86)
(part./mm) 26.9*
+ yY 2 (vol%) (vol%) 36.1 (2.86) 9.9 (0.83)
35.1 (2.50)
21.4
9.9 (0.66)
35.1 (2.50)
II
5.4 (0.61)
43.0 (2.43)
16.1
3.6 (0.37)
19.0 (2.02)
PH
7.1 (0.53)
43.5 (1.97)
26.7
1.2 (0.23)
18.6 (1.97)
NT
~
0
34.6 (2.71)
17.8
10.7 (0.75)
34.6 (2.71)
SS
-
0
41.5 (3.13)
14.0
10.0 (0.74)
41.5 (3.13)
2.1 (0.17)
19.7 (2.13)
NZ
6.1 (0.60)
41.0 (2.13)
24.7
DS
9.1 (0.77)
45.6 (2.32)
21.0
40.1 (2.43)
6.8
SP
-
0
0.0 9.8 (0.81)
23.5 (2.32) 40.1 (2.43)
80
20.5 (1.25)
53.2 (2.12)
42.3
0.0
32.6 (2.12)
SY
23.6 (1.39)
58.9 (3.84)
47.5
0.0
35.3 (3.84)
TY
16.0 (1.20)
47.0 (4.01)
76.6
0.0
31.0 (4.01)
Values are volume percents and (variances) using lineal analysis. * No variances associated with these measurements. Downloaded from jdr.sagepub.com at UNIV OF MICHIGAN on June 18, 2015 For personal use only. No other uses without permission.
1398
MAHLER & ADEY
J Dent Res Nouember 1991
TABLE 4 FACTORS INVESTIGATED FOR INFLUENCE ON CREEP Code
Factor
'
Amount of rf (vol%)
-0.845
< 0.001
y EEn' Sn
Amount of y + E + E + q' (vol%)
-0.826
< 0.001
Sn in y 1 (wt%)
0.813
0.001
-0.734
0.004
r
Introduction. The significance of creep as a factor relating to the clinical performance of dental amalgams was initially shown over 20 years ago (Mahler et al., 1970) and has since been demonstrated repeatedly. Because ofthis relationship, creep is considered ofmajor importance in characterizing dental amalgam and forms a significant part ofthe American Dental Association Specification No. 1. Information relevant to this physical property would be ofvalue in understanding the basic metallurgical and mechanical behavior of amalgam, as well as the relationship of this behavior to clinical performance. Because this paper will frequently refer to the various metallic phases of dental amalgam, these phases and their descriptions appear in Table 1. By measuring the creep ofthe y, y 1, and Y2 phases separately, Espevik (1977) concluded that the creep of amalgam is primarily due to creep of the y I matrix phase. Factors that have been shown to influence creep include final Hg content of the set amalgam (Mahler and Van Eysden, 1969), Sn in the y I phase (Mahler and Adey, 1979), Zn in the y1 phase (Johnson and Paffenbarger, 1980), grain size of y 1 (Mahleret al., 1977), amount ofthe Y2 phase (Mahler et al., 1977; Osborne and Gale, 1990), and the amount ofthe y phase (Mahler et al., 1977). In addition, the presence of small pe crystals in the y 1 matrix has been suggested as having an influence on creep (Okabe et al., 1978). Other factors that could prove to be relevant are the volume percent of n', because the amount of this phase may be related to the completeness of elimination of y2, and the total volume percent of all of the hard particles that would inhibit the creep of y 1, i.e., y + E + E + p'. In previous studies, the influence of some of these factors has been demonstrated individually or for individual alloys under various conditions. Because of this selective approach, it is not known which factors bear a relationship to each other or which factors are most significant in influencing creep. The purpose of this study was to measure all of these factors on a group of representative dental amalgams that exhibit a wide range of creep values. This would be followed by a correlation analysis and a stepwise multiple linear regression analysis, for determination of the relationships among factors and the factors that are most significant in influencing creep.
Received for publication February 1, 1991 Accepted for publication May 23, 1991 This investigation was supported by USPHS Research Grant DE 02936 from the National Institute of Dental Research, National Institutes of Health, Bethesda MD 20892.
1394
TABLE 1 GREEK LETTER DESIGNATION FOR AMALGAM PHASES y
Ag-Sn phase in the amalgam alloy (Ag3Sn, but may also include the B Ag-Sn phase)
Y1
Ag-Hg reaction matrix phase (Ag22SnHg27)
Y2
Sn-Hg reaction phase (Sn8Hg)
E
Cu-Sn phase in the amalgam alloy (Cu3Sn)
A'
Cu-Sn reaction phase (Cu6Sn5)
E Ag-Cu eutectic (approximate) Note: The formulations expressed in parentheses demonstrate the major elements present in these phases and their approximate compositions.
Materials and methods. Twelve amalgam alloys were investigated and are listed in Table 2 together with their mixing conditions. Because ofthe influence of age and batch variation, mixing conditions were determined by subjective tests for adequate trituration and mix plasticity and may not conform to the manufacturers' instructions. The volume percent of Y2 and the creep values ofthe amalgams made from these alloys, as determined in this study, are also listed in Table 2. The methods used to determine these two characteristics will be described subsequently. These alloys were selected to include different types of amalgams, as indicated by the volume percent of Y2, as well as those having different creep values. In Table 2, five amalgams can be classified as low-copper y 2-containing, four amalgams as highcopper y2-free, and three amalgams as high-copper containing small amounts of Y2 ranging from 0.7 to 3.6 volume percent. The particles for the 12 amalgam alloys were lathe-cut for four alloys, spherical for four alloys, and admixtures of lathe-cut and spherical for four alloys. Thus, the materials selected for study were reasonably representative of dental amalgam alloys, and their amalgams demonstrated a wide range of creep values. In the stepwise regression, creep was the dependent variable, and the factors investigated for influence on creep were the independent variables (IV's). The factors investigated were selected on the basis of their physical relevance to the mechanism of creep, as well as the results of the studies described in the "Introduction". In Table 4, nine selected factors are listed with their respective code designations, which will be used in the "Results" and "Discussion" sections. Also listed in this Table are the individual correlation coefficients of each of these factors to creep and their corresponding significance levels. Creep was determined in accordance with the American Dental Association Specification No. 1 for Alloy for Dental Amalgam. However, to ensure more stability in the phase structure, two weeks of storage at 370C was used, rather than the one week of the Specification, and the tests for all factors were conducted on.twoweek-old specimens. For creep, five replicate tests were run and averaged. Because the creep values ranged over three orders of
Downloaded from jdr.sagepub.com at UNIV OF MICHIGAN on June 18, 2015 For personal use only. No other uses without permission.
CREEP OF AMALGAM
Vol. 70 No. 11
1395
TABLE 2
ALLOYS, MIXING CONDITIONS, TYPE, AND CREEP
Mixing Conditions
Type
Code
Alloy Batch No.
Manufacturer
%Hg
Trituration*
y2 (vol%)
Creep (%)
AR
Aristaloy 0511781
Baker Dental, Carteret, NJ
55.0
M10
9.9
3.765
OP
Optaloy 30F70
L.D. Caulk, Milford, DE
54.0
M9
9.9
3.369
II
Optaloy II 060176
L.D. Caulk
52.5
M9
3.6
1.772
PH
Phasealloy 423/062983
Phasealloy, El Cajon, CA
51.5
M13
1.2
1.610
NT
N.T. Dentalloy 120475
S.S. White, Philadelphia, PA
49.5
M9
10.7
1.532
SS
Spherical Alloy 297501
Shofu Dental, Menlo Park, CA
44.5
M7
10.0
1.207
NZ
Dispersalloy, non-Zn 8042 J & J Ltd, Montreal, Canada
52.5
M13
0.7
1.014
DS
Dispersalloy 31426
J & J Dent, E. Windsor, NJ
51.5
M12
0.0
0.674
SP
Spheraloy T155
Kerr Mfg, Romulus, MI
47.0
M11**
9.8
0.523
80
Epoque 80 7-217/82.03
Scania Dental, Knivsta, Sweden
52.5
M7
0.0
0.222
SY
Sybraloy 1163
Kerr Mfg
48.0
H9
0.0
0.095
M9** 0.083 0.0 44.0 S.S. White TY Tytin 2038403 * Alloys mixed in a re-usable capsule with pestle (J & J Dent. Prod.) using a Varimix III amalgamator (L.D. Caulk Co.). Amalgamator speed setting and time in seconds are indicated. ** Alloys mixed as above but without a pestle.
magnitude, the natural log of 100 times the creep average was used as the dependent variable in the regression. Specimen preparation and the grinding and polishing procedure for SEM (Model JSM-T330A, JEOL USA Inc., Peabody, MA) and microprobe analysis (Model EMX-SM, Applied Research Laboratories, Sunland, CA) consisted of the following: Specimens measuring 4 mm (diam.) x 8 mm (length) were made according to ADA Specification No. 1. The specimens were mounted in brass rings with conductive epoxy resin and ground and polished in the following sequence: silicon carbide papers, 320-, 400-, and 600-grit; diamond paste, 15-, 9-, 3-, 1-, and 1/4-pum particle size. The grain size of y1 was measured on SEM photographs of etched amalgams at magnifications of 5000 x and 10,000 x, depending on the sizes of the grains. Etching was done with solution No. 1 of Wings etch (Wing, 1965) at 1/10 strength in a Minimet polisher on a microcloth (Buehler Ltd., Lake Bluff, IL), with a pressure setting of zero. A sheet of acetate paper with evenly spaced lines was placed over the SEM photographs, and the numbers of grains lying beneath fixed line lengths were counted and converted to an average grain size in pm. From 50 to 90 line lengths were utilized to produce an overall average grain size. Although this method underestimates the true grain size, the difference is the same for all alloys and would not be of consequence when used in a regression analysis. Sn in the y 1 phase was measured using the following microprobe settings: excitation potential = 15 keV; sample current = 5 nA; counting time = 60 s; analyzed lines - Ag La, Hg Mcx, Sn Lc, and Cu Kca. For determination of weight percent, the raw counts were corrected for atomic number, absorption, and fluorescence effects (ZAF) by use of the program by Colby (1968) known as MAGIC IV. Zn content ofthe amalgam alloy was measured by flame atomic absorption spectrophotometry(Model IL 151, Instrumentation Laboratories, Inc., Wilmington, MA). The accuracy of measurement was + 0.0 1% Zn, and the average ofthree replicate specimens was used.
The Hg content of the set amalgam was determined by the method of Crawford and Larson (1955). The five specimens tested for creep were used for determination of the average weight percent of Hg. All of the variables indicating the amount of each phase in the amalgam microstructure were measured by x-ray line profiles using lineal analysis (DeHoff and Rhines, 1968). The microprobe beam was kept stationary, and the specimen moved under the beam at a rate of 1.3 pm/s. A four-pen recorder registered the simultaneous signals of the Ag Lcx, Hg Mcx, Sn La, and Cu KaL lines. The individual phases were identified by the appropriate line peaks, and the width of the best element line for each phase was measured at 1/2 the maximum peak height. For determination of the volume percent ofeach phase, all ofthe calculated widths were summed and divided by the total distance traveled on the specimen. A distance of from 4 to 7 mm was traced on each specimen. Variances for this lineal analysis were determined by the method described by Hilliard and Cahn (1961). For determination of the number of small q' crystals per mm, Cu peaks whose widths at 1/2 their peak heights corresponded to less than 1.5 pm in actual size were counted and divided by the total distance traversed. There was no variance associated with this measurement. Performance of stepwise regression is considered tenuous when dealing with a large number of independent variables and a relatively small sample size (Cohen and Cohen, 1983). Therefore, the following procedure was used to reduce the number of independent variables to be considered in the stepwise regression: (1) A correlation analysis was performed among the independent variables (IVs). (2) From this analysis, the IV's were separated into two groups. Group A included the IV's which were not highly correlated with any other IV. Group B included the IVs which were all highly correlated with each other (Table 5). (3) Separate, stepwise, multiple linear regressions were then
Downloaded from jdr.sagepub.com at UNIV OF MICHIGAN on June 18, 2015 For personal use only. No other uses without permission.
J Dent Res November 1991
MAHLER & ADEY
1396
0
R2= 0.949
w
r=
a-
-0.845
U,
W
w U
a-
44
w LL
0 0
x0
z -i
0
3
*
z
+
LOW Cu
-I
10%Y2
® HIGH Cu I-44%Y2
n1
2 0
HIGH Cu
LOW Cu
10%a2
- 4% 2 HIGH Cu 0%0'2
(3 HIGH Cu
W
°%'Y2
I
I
5
10
I 15
I
20
25 LN 100
7'- VOLUME %
Fig. 1-Measured creep vs. volume percent of p'.
x
CREEP
-
CALCULATED
Fig. 2-Measured creep vs. creep calculated from the regression equation.
in Table 5. Based on these coefficients, the independent variables separated into the two groups, A and B, shown in this Table. With two exceptions, Group A lists the IV's not highly correlated with any other IV (p 0.05), while Group B lists the IVs that were highly correlated to each other (p < 0.001). In Table 6, the results of the trial stepwise multiple linear regressions are presented. The column labeled "Starting Matrix" shows the variables used for initiation of the regression procedure. Results. Group A variables by themselves were run first. Subsequently, all Tables 3A and 3B list the data for the 12 alloys and nine factors of the five 's of Group A were used together with each of the IV's evaluated in order of decreasing values for creep. In Table 3A, Sn in of Group B in separate regressions. The column labeled "Variables Accepted" are those IV's of the the y 1, phase was relatively high for the y 2-containing amalgams and decreased significantly as Y2 was eliminated. The grain size of starting matrix that were accepted by the stepwise regression with yI showed significant differences, particularly for the relatively an overall a set for 0.15 (Bendel and Afifi, 1977). The column large grains of alloy SP and alloy TY. The Hg contents ofthe various labeled "Adjusted Re" indicates the R2 values for the multiple linear amalgams as condensed by the method of ADA Specification No. 1 regressions of the IV's listed in the "Variables Accepted" column. are in some cases extremely low compared with what would be The term "Adjusted" means that the raw R2 values have been expected under clinically simulated condensation procedures adjusted for the number of IV's in the equation and the number of cases. This parameter is a more conservative estimate ofthe percent (Mahler, 1979). The weight percent of Zn revealed four alloys that were essen- of variance explained, especially when the sample size is small tially free of Zn (SS, NZ, SY, and TY), while the remaining eight (Cohen and Cohen, 1983). The highest RI of0.949 was achieved by use ofthe five variables: alloys showed varied Zn contents up to approximately 1%. In Table 3B, the volume percent of p was zero for the amalgams volume percent of p', the grain size of y the volume percent of y containing approximately 10% y2 and increased to values up to 24% + E, the number of small crystalsper mm, and the weight percent for the y 2-free amalgams. The volume percent of Y2 was essentially of Hg in the set amalgam. The results of this regression are detailed 10% in the five low-Cu alloys (AR, OP, NT, SS, and SP), 1-4% in three in Table 7, where the listed order of the five variables reflects the dispersed-phase alloys (II, PH, and NZ), and 0% in one dispersed- relative influence of these variables on increasing R2, i.e., their phase alloy (DS) and in the three high-Cu single- composition alloys relative ability to reduce the average variance ofthe points around the regression line. Thus, p' volume percent was the most influen(80, SY, and TY). Table 4 lists the correlation coefficients between creep and each tial factor affecting creep, y grain size was the next most influential ofthe IV's. The first four variables listed showed significant correla- factor, and so on down the column labeled "Ln Creep vs". Changes tions with creep at p < 0.01. These were the volume percent of the 9' in the adjusted R12 and the Student t and p values for each factor are phase, the total volume percent ofy + E + E + n', the weight percent presented as each IV is added to the equation. The t and corresponding ofSn in the y 1 phase, and the number of small crystalsper mm of the p values associated with each factor indicate the significance of the 9' phase. Of these factors, the one producing the highest correlation contribution of that factor to the best-fitting regression. All factors with creep proved to be the volume percent of the p' phase with a had t values significant at p < 0.05, except for the weight percent of correlation coefficient of-0.845 (significant at p < 0.001). A plot of In Hg, for which p = 0.055, slightly higher than 0.05. However, this (creep) vs. volume percent of n 'is shown in Fig. 1. [Vwas retained because without its presence, one of the test alloys Significant correlations among the IV's themselves can be seen was found to be an outlier. performed on the IV's of Group A together with each of the IV's of Group B. (4) The most significant factors influencing creep were identified as those producing a regression having the highest R12. The stepwise multiple linear regression was run by the computer program known as SYSTAT (Systat, Inc., Evanston, IL).
were
1,
1
Downloaded from jdr.sagepub.com at UNIV OF MICHIGAN on June 18, 2015 For personal use only. No other uses without permission.
1397
CREEP OF AMALGAM
Vol. 70 No. 11
TABLE 3A DATA USED IN THE REGRESSION Creep (%) n=5
Sn in y1 (Wt%) n=8
Y 1 Grain Size
Hg Content
(pm) n=75
(wt%) n=5
Zn in y (wt%) n=3
AR
3.77 (0.193)
2.79 (0.307)
1.13 (0.352)
46.7 (0.62)
0.53 (0.022)
OP
3.37 (0.460)
2.94 (0.281)
1.14 (0.449)
45.7 (0.29)
0.31 (0.009)
II
1.77 (0.380)
2.32 (0.182)
1.06 (0.426)
45.4 (0.72)
0.28 (0.017)
PH
1.61 (0.075)
1.55 (0.324)
1.32 (0.339)
44.6 (0.14)
0.24 (0.011)
NT
1.53 (0.410)
2.58 (0.145)
1.01 (0.266)
42.1 (0.37)
0.96 (0.003)
SS
1.21 (0.306)
2.46 (0.170)
1.53 (0.701)
42.9 (0.26)
0
NZ
1.01 (0.122)
2.12 (0.244)
1.53 (0.549)
46.7 (0.36)
0
DS
0.67 (0.070)
1.48 (0.345)
1.65 (1.556)
44.1 (0.25)
1.02 (0.020)
SP
0.52 (0.233)
2.37 (0.321)
2.02 (0.831)
44.2 (0.24)
0.50 (0.024)
80
0.22 (0.091)
0.98 (0.359)
1.39 (0.429)
44.5 (0.07)
0.10 (0.025)
SY
0.10 (0.038)
0.69 (0.285)
1.28 (0.499)
43.1 (0.56)
0.02 (0.013)
TY
0.08 (0.045)
1.52 (0.341)
3.49 (2.878)
40.7 (0.16)
0
Alloy
Values are means and standard deviations. TABLE 3B DATA USED IN THE REGRESSION -yEfl'
Alloy AR
OP
(vol%) - 0 -
0
(vol%) 36.1 (2.86)
(part./mm) 26.9*
+ yY 2 (vol%) (vol%) 36.1 (2.86) 9.9 (0.83)
35.1 (2.50)
21.4
9.9 (0.66)
35.1 (2.50)
II
5.4 (0.61)
43.0 (2.43)
16.1
3.6 (0.37)
19.0 (2.02)
PH
7.1 (0.53)
43.5 (1.97)
26.7
1.2 (0.23)
18.6 (1.97)
NT
~
0
34.6 (2.71)
17.8
10.7 (0.75)
34.6 (2.71)
SS
-
0
41.5 (3.13)
14.0
10.0 (0.74)
41.5 (3.13)
2.1 (0.17)
19.7 (2.13)
NZ
6.1 (0.60)
41.0 (2.13)
24.7
DS
9.1 (0.77)
45.6 (2.32)
21.0
40.1 (2.43)
6.8
SP
-
0
0.0 9.8 (0.81)
23.5 (2.32) 40.1 (2.43)
80
20.5 (1.25)
53.2 (2.12)
42.3
0.0
32.6 (2.12)
SY
23.6 (1.39)
58.9 (3.84)
47.5
0.0
35.3 (3.84)
TY
16.0 (1.20)
47.0 (4.01)
76.6
0.0
31.0 (4.01)
Values are volume percents and (variances) using lineal analysis. * No variances associated with these measurements. Downloaded from jdr.sagepub.com at UNIV OF MICHIGAN on June 18, 2015 For personal use only. No other uses without permission.
1398
MAHLER & ADEY
J Dent Res Nouember 1991
TABLE 4 FACTORS INVESTIGATED FOR INFLUENCE ON CREEP Code
Factor
'
Amount of rf (vol%)
-0.845
< 0.001
y EEn' Sn
Amount of y + E + E + q' (vol%)
-0.826
< 0.001
Sn in y 1 (wt%)
0.813
0.001
-0.734
0.004
r
